Purpose: To demonstrate how budgetary constraints may alter a radiologist’s diagnostic decisions in the pursuit of optimal breast cancer diagnosis as measured by quality adjusted life years (QALYs) from the societal perspective.
Method: We developed a finite-horizon discrete-time Markov decision process (MDP) model to optimize breast cancer diagnosis problem i.e. given the demographic data and mammography features, what is the optimal course of action; routine screening, short-term follow-up or biopsy? The MDP model uses breast cancer progression probabilities obtained from 62,219 consecutive mammography records reported in Breast Imaging Reporting and Data System (BI-RADS) format. We modify a linear programming formulation of the MDP model to include budgetary restrictions and solve it to maximize the total expected QALYs of a patient. We repeat this experiment for various budget values in a feasible range. We compare actual clinical practice with optimal decisions obtained using the model and conduct incremental cost effectiveness analysis over the range identified.
Result: We find that the optimal policies are of double-control-limit type i.e. for each age there exists a certain probability of cancer over which the optimal decision is short-term follow-up and a secondary threshold over which the optimal decision is biopsy. Under budgetary restrictions, initially short-term follow-ups are reduced and as the budget gets tighter, the optimal threshold to biopsy increases. The effect of budget in short-term follow-ups and the biopsy thresholds are more pronounced in older ages. Moreover, diagnostic decisions based on age and breast cancer risk while observing optimal thresholds result in cost savings without sacrificing QALYs. Comparing to actual clinical practice, using optimal thresholds for decision-making may result in approximately 14% cost savings without sacrificing QALYs.
Conclusion: In this work, we present a novel framework for evaluating the cost-effectiveness of diagnostic procedures in the context of a sequential decision making problem, namely breast cancer diagnostic decisions after mammography. In conducting our analysis, we only consider maximizing QALYs and consider cost only as a restriction for optimal actions. Under this framework, our analysis indicate that short-term follow ups are the immediate target when budget becomes a concern.
Candidate for the Lee B. Lusted Student Prize Competition